Neutral Substrates: A Design Constraint for Shared Records Under Persistent Interpretive Disagreement

Technical Summary: Neutral Substrates for Shared Records Under Persistent Interpretive Disagreement

Problem Setting
Shared accountability records are frequently utilized by diverse parties (e.g., institutions, auditors, regulators, affected individuals) who may share a common reference to an event, entity, or decision but hold persistent, potentially permanent disagreements regarding causation, responsibility, justification, or normative interpretation.

The paper identifies a structural design failure in such records: a shared record cannot remain neutral if its foundational layer makes an "object-level substrate-layer commitment" to a contested causal or normative proposition. If the record asserts a specific interpretation (e.g., "Event A caused Event B" or "Decision D was justified by Rule R") as a fact of the substrate itself, it structurally aligns with one interpretive framework against others. Conversely, omitting contested claims entirely suppresses the information necessary for accountability. The core problem is how to preserve contested claims without adopting them as the shared base's truth.

Methodology and Definitions
The paper employs standard representational machinery—reification, attribution, and provenance—within a formal ontology framework. It does not introduce new logical mechanisms but applies them to define a specific design constraint.

Key definitions include:

• Substrate ($S$): A shared representational base providing stable reference (individuation, co-reference, persistence) for entities and artifacts across a class of admissible interpretive frameworks ($F$).
• Substrate-Layer Commitment: A proposition $p$ is committed by the substrate if $S \vdash p$ independently of any framework.
• Attribution Proposition: A proposition of the form $Asserts(x, \phi)$, indicating that a specific source $x$ asserts $\phi$. The substrate commits to the attribution, not the content $\phi$.
• Object-Level Interpretive Proposition: The asserted proposition $\phi$ itself.
• Admissible Framework: A framework satisfying internal consistency, evidentiary grounding, and a documented interpretive function. Admissibility does not imply agreement; frameworks may disagree on causation or norms.
• Referential Common Ground: An assumption that the referential commitments ($S_{ref}$) of the substrate are compatible with every admissible framework.

The Neutrality Constraint
The paper's central contribution is a biconditional design constraint, Constraint 5.2 (Neutrality), which guarantees "neutrality by design" under two specific assumptions:

1. Contestability: Causal and normative propositions are not guaranteed to be framework-invariant at design time.
2. Referential Common Ground: The substrate's referential commitments are compatible with all admissible frameworks.

The Constraint:
A substrate $S$ is neutral by design (guaranteed to satisfy interpretive non-commitment and extension stability) if and only if its foundational layer is restricted to:

1. Referential commitments ($S_{ref}$): Identifiers, types, timestamps, provenance, and relations fixing what is being referred to.
2. Permitted attribution propositions: Assertions of the form $Asserts(x, \phi)$ where the attributional basis (source, occurrence, provenance) is fixed by $S_{ref}$.

Crucially, the substrate must not make any object-level causal or normative commitments. Causal and normative content may exist in the record, but only as the content of attributed assertions (e.g., "Framework F asserts that X caused Y"), not as the substrate's own assertion that X caused Y.

Mechanism and Results
The mechanism for achieving neutrality is reification. Instead of asserting a contested proposition $\phi$, the substrate asserts $Asserts(x, \phi)$.

• Sufficiency: If the foundational layer is restricted to referential commitments and permitted attributions, the substrate remains compatible with every admissible framework (Extension Stability) and does not commit to disputed interpretations (Interpretive Non-Commitment).
• Necessity: If the substrate commits to any proposition outside these permitted classes (specifically, an object-level causal or normative proposition), neutrality by design cannot be guaranteed because the assumptions do not certify its compatibility with all frameworks.

Boundary Conditions
The paper explicitly defines the limits of this constraint. Neutrality is unavailable if:

• Admissible frameworks contest the substrate's referential regime (e.g., disagreeing on whether two records refer to the same event or if an identity persists through a merger).
• Admissible frameworks contest the attributional basis (e.g., disagreeing on who made an assertion, what was asserted, or the provenance of the assertion).
  In these boundary cases, the disagreement lies in the reference or attribution itself, not just the interpretation, preventing the establishment of a neutral shared base.

Significance and Claims
The paper claims to provide a checkable condition for the foundational layer of shared accountability records. Its significance lies in:

1. Formalizing Neutrality: Moving beyond the idea that neutrality requires omitting content, it defines neutrality as the separation of attribution from assertion at the foundational layer.
2. Design-Time Guarantee: It offers a structural guarantee that a record will remain usable by future, unknown admissible frameworks, provided the referential common ground holds.
3. Preserving Accountability: It allows records to capture "who said what" and the provenance of claims without forcing the shared system to adjudicate the truth of those claims.

The paper explicitly states it does not:

• Propose a specific data format (it applies to relational, graph, RDF, etc.).
• Adjudicate which framework is correct.
• Solve general accountability or determine liability.
• Claim that reference is always shared (it identifies the boundary where reference is contested).

The work serves as a structural condition for records under persistent disagreement, ensuring that the shared substrate supports disagreement without erasing contested claims or adopting them as the system's truth.